To notify userspace about which task (if any) made the device get in a
wedge state, make use of drm_wedge_task_info parameter, filling it with
the task PID and name.
Signed-off-by: André Almeida <andrealmeid@igalia.com>
---
v8:
- Drop check before calling amdgpu_vm_put_task_info()
- Drop local variable `info`
v7:
- Remove struct cast, now we can use `info = &ti->task`
- Fix struct lifetime, move amdgpu_vm_put_task_info() after
drm_dev_wedged_event() call
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 13 +++++++++++--
drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 7 +++++--
2 files changed, 16 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 8a0f36f33f13..a59f194e3360 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -6363,8 +6363,17 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
atomic_set(&adev->reset_domain->reset_res, r);
- if (!r)
- drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE, NULL);
+ if (!r) {
+ struct amdgpu_task_info *ti = NULL;
+
+ if (job)
+ ti = amdgpu_vm_get_task_info_pasid(adev, job->pasid);
+
+ drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE,
+ ti ? &ti->task : NULL);
+
+ amdgpu_vm_put_task_info(ti);
+ }
return r;
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
index 0c1381b527fe..1e24590ae144 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
@@ -89,6 +89,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
{
struct amdgpu_ring *ring = to_amdgpu_ring(s_job->sched);
struct amdgpu_job *job = to_amdgpu_job(s_job);
+ struct drm_wedge_task_info *info = NULL;
struct amdgpu_task_info *ti;
struct amdgpu_device *adev = ring->adev;
int idx;
@@ -125,7 +126,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
ti = amdgpu_vm_get_task_info_pasid(ring->adev, job->pasid);
if (ti) {
amdgpu_vm_print_task_info(adev, ti);
- amdgpu_vm_put_task_info(ti);
+ info = &ti->task;
}
/* attempt a per ring reset */
@@ -164,13 +165,15 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
if (amdgpu_ring_sched_ready(ring))
drm_sched_start(&ring->sched, 0);
dev_err(adev->dev, "Ring %s reset succeeded\n", ring->sched.name);
- drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE, NULL);
+ drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE, info);
goto exit;
}
dev_err(adev->dev, "Ring %s reset failure\n", ring->sched.name);
}
dma_fence_set_error(&s_job->s_fence->finished, -ETIME);
+ amdgpu_vm_put_task_info(ti);
+
if (amdgpu_device_should_recover_gpu(ring->adev)) {
struct amdgpu_reset_context reset_context;
memset(&reset_context, 0, sizeof(reset_context));
--
2.49.0
On 6/17/25 14:49, André Almeida wrote:
> To notify userspace about which task (if any) made the device get in a
> wedge state, make use of drm_wedge_task_info parameter, filling it with
> the task PID and name.
>
> Signed-off-by: André Almeida <andrealmeid@igalia.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Do you have commit right for drm-misc-next?
Regards,
Christian.
> ---
> v8:
> - Drop check before calling amdgpu_vm_put_task_info()
> - Drop local variable `info`
> v7:
> - Remove struct cast, now we can use `info = &ti->task`
> - Fix struct lifetime, move amdgpu_vm_put_task_info() after
> drm_dev_wedged_event() call
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 13 +++++++++++--
> drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 7 +++++--
> 2 files changed, 16 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> index 8a0f36f33f13..a59f194e3360 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> @@ -6363,8 +6363,17 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
>
> atomic_set(&adev->reset_domain->reset_res, r);
>
> - if (!r)
> - drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE, NULL);
> + if (!r) {
> + struct amdgpu_task_info *ti = NULL;
> +
> + if (job)
> + ti = amdgpu_vm_get_task_info_pasid(adev, job->pasid);
> +
> + drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE,
> + ti ? &ti->task : NULL);
> +
> + amdgpu_vm_put_task_info(ti);
> + }
>
> return r;
> }
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
> index 0c1381b527fe..1e24590ae144 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
> @@ -89,6 +89,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
> {
> struct amdgpu_ring *ring = to_amdgpu_ring(s_job->sched);
> struct amdgpu_job *job = to_amdgpu_job(s_job);
> + struct drm_wedge_task_info *info = NULL;
> struct amdgpu_task_info *ti;
> struct amdgpu_device *adev = ring->adev;
> int idx;
> @@ -125,7 +126,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
> ti = amdgpu_vm_get_task_info_pasid(ring->adev, job->pasid);
> if (ti) {
> amdgpu_vm_print_task_info(adev, ti);
> - amdgpu_vm_put_task_info(ti);
> + info = &ti->task;
> }
>
> /* attempt a per ring reset */
> @@ -164,13 +165,15 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
> if (amdgpu_ring_sched_ready(ring))
> drm_sched_start(&ring->sched, 0);
> dev_err(adev->dev, "Ring %s reset succeeded\n", ring->sched.name);
> - drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE, NULL);
> + drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE, info);
> goto exit;
> }
> dev_err(adev->dev, "Ring %s reset failure\n", ring->sched.name);
> }
> dma_fence_set_error(&s_job->s_fence->finished, -ETIME);
>
> + amdgpu_vm_put_task_info(ti);
> +
> if (amdgpu_device_should_recover_gpu(ring->adev)) {
> struct amdgpu_reset_context reset_context;
> memset(&reset_context, 0, sizeof(reset_context));
Em 17/06/2025 10:07, Christian König escreveu: > On 6/17/25 14:49, André Almeida wrote: >> To notify userspace about which task (if any) made the device get in a >> wedge state, make use of drm_wedge_task_info parameter, filling it with >> the task PID and name. >> >> Signed-off-by: André Almeida <andrealmeid@igalia.com> > > Reviewed-by: Christian König <christian.koenig@amd.com> > > Do you have commit right for drm-misc-next? > I've merged the series into drm-misc-next. Thanks! André
Em 17/06/2025 10:07, Christian König escreveu:
> On 6/17/25 14:49, André Almeida wrote:
>> To notify userspace about which task (if any) made the device get in a
>> wedge state, make use of drm_wedge_task_info parameter, filling it with
>> the task PID and name.
>>
>> Signed-off-by: André Almeida <andrealmeid@igalia.com>
>
> Reviewed-by: Christian König <christian.koenig@amd.com>
>
> Do you have commit right for drm-misc-next?
>
Thanks for the reviews!
I do have access, but if you don't mind, can you push this one?
> Regards,
> Christian.
>
>> ---
>> v8:
>> - Drop check before calling amdgpu_vm_put_task_info()
>> - Drop local variable `info`
>> v7:
>> - Remove struct cast, now we can use `info = &ti->task`
>> - Fix struct lifetime, move amdgpu_vm_put_task_info() after
>> drm_dev_wedged_event() call
>> ---
>> drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 13 +++++++++++--
>> drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 7 +++++--
>> 2 files changed, 16 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> index 8a0f36f33f13..a59f194e3360 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> @@ -6363,8 +6363,17 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
>>
>> atomic_set(&adev->reset_domain->reset_res, r);
>>
>> - if (!r)
>> - drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE, NULL);
>> + if (!r) {
>> + struct amdgpu_task_info *ti = NULL;
>> +
>> + if (job)
>> + ti = amdgpu_vm_get_task_info_pasid(adev, job->pasid);
>> +
>> + drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE,
>> + ti ? &ti->task : NULL);
>> +
>> + amdgpu_vm_put_task_info(ti);
>> + }
>>
>> return r;
>> }
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
>> index 0c1381b527fe..1e24590ae144 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
>> @@ -89,6 +89,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
>> {
>> struct amdgpu_ring *ring = to_amdgpu_ring(s_job->sched);
>> struct amdgpu_job *job = to_amdgpu_job(s_job);
>> + struct drm_wedge_task_info *info = NULL;
>> struct amdgpu_task_info *ti;
>> struct amdgpu_device *adev = ring->adev;
>> int idx;
>> @@ -125,7 +126,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
>> ti = amdgpu_vm_get_task_info_pasid(ring->adev, job->pasid);
>> if (ti) {
>> amdgpu_vm_print_task_info(adev, ti);
>> - amdgpu_vm_put_task_info(ti);
>> + info = &ti->task;
>> }
>>
>> /* attempt a per ring reset */
>> @@ -164,13 +165,15 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
>> if (amdgpu_ring_sched_ready(ring))
>> drm_sched_start(&ring->sched, 0);
>> dev_err(adev->dev, "Ring %s reset succeeded\n", ring->sched.name);
>> - drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE, NULL);
>> + drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE, info);
>> goto exit;
>> }
>> dev_err(adev->dev, "Ring %s reset failure\n", ring->sched.name);
>> }
>> dma_fence_set_error(&s_job->s_fence->finished, -ETIME);
>>
>> + amdgpu_vm_put_task_info(ti);
>> +
>> if (amdgpu_device_should_recover_gpu(ring->adev)) {
>> struct amdgpu_reset_context reset_context;
>> memset(&reset_context, 0, sizeof(reset_context));
>
On 6/17/25 15:22, André Almeida wrote:
> Em 17/06/2025 10:07, Christian König escreveu:
>> On 6/17/25 14:49, André Almeida wrote:
>>> To notify userspace about which task (if any) made the device get in a
>>> wedge state, make use of drm_wedge_task_info parameter, filling it with
>>> the task PID and name.
>>>
>>> Signed-off-by: André Almeida <andrealmeid@igalia.com>
>>
>> Reviewed-by: Christian König <christian.koenig@amd.com>
>>
>> Do you have commit right for drm-misc-next?
>>
>
> Thanks for the reviews!
>
> I do have access, but if you don't mind, can you push this one?
Sure, but give me till the end of today.
(And maybe ping me next week should I forget about it).
Regards,
Christian.
>
>> Regards,
>> Christian.
>>
>>> ---
>>> v8:
>>> - Drop check before calling amdgpu_vm_put_task_info()
>>> - Drop local variable `info`
>>> v7:
>>> - Remove struct cast, now we can use `info = &ti->task`
>>> - Fix struct lifetime, move amdgpu_vm_put_task_info() after
>>> drm_dev_wedged_event() call
>>> ---
>>> drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 13 +++++++++++--
>>> drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 7 +++++--
>>> 2 files changed, 16 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>>> index 8a0f36f33f13..a59f194e3360 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>>> @@ -6363,8 +6363,17 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
>>> atomic_set(&adev->reset_domain->reset_res, r);
>>> - if (!r)
>>> - drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE, NULL);
>>> + if (!r) {
>>> + struct amdgpu_task_info *ti = NULL;
>>> +
>>> + if (job)
>>> + ti = amdgpu_vm_get_task_info_pasid(adev, job->pasid);
>>> +
>>> + drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE,
>>> + ti ? &ti->task : NULL);
>>> +
>>> + amdgpu_vm_put_task_info(ti);
>>> + }
>>> return r;
>>> }
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
>>> index 0c1381b527fe..1e24590ae144 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
>>> @@ -89,6 +89,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
>>> {
>>> struct amdgpu_ring *ring = to_amdgpu_ring(s_job->sched);
>>> struct amdgpu_job *job = to_amdgpu_job(s_job);
>>> + struct drm_wedge_task_info *info = NULL;
>>> struct amdgpu_task_info *ti;
>>> struct amdgpu_device *adev = ring->adev;
>>> int idx;
>>> @@ -125,7 +126,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
>>> ti = amdgpu_vm_get_task_info_pasid(ring->adev, job->pasid);
>>> if (ti) {
>>> amdgpu_vm_print_task_info(adev, ti);
>>> - amdgpu_vm_put_task_info(ti);
>>> + info = &ti->task;
>>> }
>>> /* attempt a per ring reset */
>>> @@ -164,13 +165,15 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
>>> if (amdgpu_ring_sched_ready(ring))
>>> drm_sched_start(&ring->sched, 0);
>>> dev_err(adev->dev, "Ring %s reset succeeded\n", ring->sched.name);
>>> - drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE, NULL);
>>> + drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE, info);
>>> goto exit;
>>> }
>>> dev_err(adev->dev, "Ring %s reset failure\n", ring->sched.name);
>>> }
>>> dma_fence_set_error(&s_job->s_fence->finished, -ETIME);
>>> + amdgpu_vm_put_task_info(ti);
>>> +
>>> if (amdgpu_device_should_recover_gpu(ring->adev)) {
>>> struct amdgpu_reset_context reset_context;
>>> memset(&reset_context, 0, sizeof(reset_context));
>>
>
Hi Christian, Em 18/06/2025 04:29, Christian König escreveu: > On 6/17/25 15:22, André Almeida wrote: >> Em 17/06/2025 10:07, Christian König escreveu: >>> On 6/17/25 14:49, André Almeida wrote: >>>> To notify userspace about which task (if any) made the device get in a >>>> wedge state, make use of drm_wedge_task_info parameter, filling it with >>>> the task PID and name. >>>> >>>> Signed-off-by: André Almeida <andrealmeid@igalia.com> >>> >>> Reviewed-by: Christian König <christian.koenig@amd.com> >>> >>> Do you have commit right for drm-misc-next? >>> >> >> Thanks for the reviews! >> >> I do have access, but if you don't mind, can you push this one? > > Sure, but give me till the end of today. > It was already merged, no worries!
Em 17/06/2025 10:22, André Almeida escreveu: > Em 17/06/2025 10:07, Christian König escreveu: >> On 6/17/25 14:49, André Almeida wrote: >>> To notify userspace about which task (if any) made the device get in a >>> wedge state, make use of drm_wedge_task_info parameter, filling it with >>> the task PID and name. >>> >>> Signed-off-by: André Almeida <andrealmeid@igalia.com> >> >> Reviewed-by: Christian König <christian.koenig@amd.com> >> >> Do you have commit right for drm-misc-next? >> > > Thanks for the reviews! > > I do have access, but if you don't mind, can you push this one? > Never mind, I can push this one myself :)
© 2016 - 2025 Red Hat, Inc.